When a Service doesn't answer, Kubernetes rarely prints an error. Work through the chain in order.
First, the endpoints: kubectl get endpoints NAME. If the list is empty, the selector matches no Ready Pod. Compare the selector in kubectl describe service with kubectl get pods --show-labels; one character is enough. If the labels match but the Pods show READY 0/1, their readiness probe fails, and kubectl describe pod shows the Unhealthy events.
If there are endpoints but calls are refused, the traffic goes to the wrong port. targetPort must be a port the container listens on, given as a number or as the name of one of the container's ports.
Last, check the selector isn't too broad: a Service sends traffic to every Ready Pod that carries its labels, including a canary or a debug copy someone started with the same app label.
Endpoints, selector, labels, readiness: in that order.